magicRuns for warpwarp template exposures sometimes don't get inverse set

[bills]

Some fraction of the time magicRun's based on warp-warp diffs are getting inserted incorrectly.

The magicRun for a template exposure should have magicRun.inverse set to 1.

Sometimes this does not happen.
For example 100944, 45, 47, 49, 50, 52, 56, and 58.

When this happens everything gets processed all the way to the
end and then processing of the root node fails.

The examples above were changed by hand. Then the runs completed.

[bills]

The bug was in the sql.

CONVERT(SUBSTRING_INDEX(GROUP_CONCAT(diffWarps.inverse ORDER BY diffWarps.diff_id), ',', 1), UNSIGNED) AS inverse,

Needed to be order by diff_id DESC

CONVERT(SUBSTRING_INDEX(GROUP_CONCAT(diffWarps.inverse ORDER BY diffWarps.diff_id DESC), ',', 1), UNSIGNED) AS inverse,

Fixed in r29831.